- Old dependencies R-packages were removed
- Calls to functions ‘Kronecker’ and ‘Kronecker_cov’ from the tensorEVD R-package were added
- Functions to work with triangular matrices were removed

**New features**

- Function ‘fitBLUP’ allows the solution of the mixed model for multiple traits when input ‘y’ has more than one column
- Function ‘getGenCov’ allows the calculation of all pairs of columns of input ‘y’ so a genetic covariance matrix can be formed
- Function ‘SSI’ is extended to the multi-trait case if input ‘y’ has more than one column. In this case within-trait genetic/residual covariances varU and varE are calculated using function ‘getGenCov’ when are not provided
- Arguments ‘trn’ and ‘tst’ in function ‘SSI(…,trn,tst)’ can be be now passed as ‘SSI(…,trn_tst)’, where ‘trn_tst’ can be a vector with 0’s (for tst) and 1’s (for trn)
- New function ‘prune’ added (see manual)
- New functions to work with triangular matrices added (see manual)

**New features**

- Functions ‘solveEN’ and ‘LARS’ allow solving several regressions by iterating over columns of argument ‘Gamma’
- Function ‘SSI’ allows either saving or returning the coefficients through ‘save.beta’ and ‘return.beta’ arguments
- Function ‘SSI’ returns also genetic values ‘u’ of testing subjects
- Methods ‘summary’, ‘fitted’, and ‘plot’ can be implemented for a desired response variable ‘y’ different from the specified in object$y, e.g., fitted(object, y)

**New features**

- Some problems were fixed in the documentation structure as required by CRAN.
- Functions ‘lars2’, ‘SSI_CV’, ‘plotNet’, ‘plotPath’ changed their names to ‘LARS’, ‘SSI.CV’, ‘net.plot’, and ‘path.plot’, respectively.
- Some arguments’ functions changed their names to a more informative name (e.g., ‘minLambda’ => ‘lambda.min’)
- More functionalities added to ‘net.plot’ function

**New features**

- Functions ‘SSI’ and ‘SSI_CV’ allow providing either ‘theta’ (residual/genetic variances ratio) or the ‘h2’ (heritability)

**Bug fixes**

- C-based routine associated to the ‘readBinary’ function now uses the ‘Rf_allocMatrix’ method to handle matrices whose length (number of rows x number of columns) exceed 2^31-1 = 2147483647

**New features**

- Function ‘solveEN’ allows early stop when a user-provided number of non-zero predictors (at a given value of lambda) is reached (argument ‘maxDF’)
- Functions ‘solveEN’ and ‘lars2’ return object ‘beta’ as matrix with predictors in rows (rather than in columns)
- Function ‘cov2cor2’ allows multiplying the resulting correlation matrix times a constant ‘a’ (default is ‘a=1’)
- Provided ‘wheatHTP’ dataset includes now an array of 4-folds partitions (‘CV’ column in object ‘Y’) and calculations of genetic and residual covariances between YLD and each of the wavelengths (‘genCOV_xy’ and ‘resCOV_xy’ objects), and among YLD from each environment (‘genCOV_yy’ object). Residuals covariances among YLD from each environment (‘resCOV_yy’ object) is also included

**Bug fixes**

- Function ‘fitBLUP’ performs the new checking varU <= 2*var(y) to declare a possible error if FALSE
- Function reshape2::melt is used instead of reshape::melt

**New features**

- More detailed functions’ documentation
- Function ‘fitBLUP’ performs a quality control for very small or negatives eigenvalues
- Function ‘saveBinary’ does not save columns’ nor rows’ names anymore
- Function ‘SSI’ uses now a C-based routine called ‘add2diag’ created to add a numeric value to the diagonal of a symmetric matrix (single or double precision). This routine is not at the user level
- Function ‘getGenCov’ has the argument ‘warn’ to whether show warnings from ‘BLUP’ analyses

**Bug fixes**

- All C-based routines: a ‘long long’ variable type, instead of an ‘int’ type, was used for indexing arrays (matrices). This change allows dealing with matrices whose length (number of rows x number of columns) exceed 2^31-1 = 2147483647 (e.g., a matrix of 46341 x 46341)

**Features**

- First released version
- Function ‘solveMixed’ (from GitHub version) was renamed to ‘fitBLUP’